Llama0-3-8b-v0.1-p-2-lr5e-6-e1

This model is a fine-tuned version of meta-llama/Meta-Llama-3-8B-Instruct on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.0071
• Rewards/chosen: -325.0106
• Rewards/rejected: -326.4964
• Rewards/accuracies: 0.4839
• Rewards/margins: 1.4858
• Logps/rejected: -32736.4023
• Logps/chosen: -32589.3574
• Logits/rejected: 15.2095
• Logits/chosen: 15.1707

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-06
• train_batch_size: 2
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• gradient_accumulation_steps: 8
• total_train_batch_size: 128
• total_eval_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 1.0

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.0157	0.2137	100	0.0113	-210.0611	-211.0992	0.4597	1.0381	-21196.6719	-21094.4004	13.2797	13.2492
0.0103	0.4275	200	0.0100	-295.4855	-293.2994	0.5081	-2.1861	-29416.6973	-29636.8457	14.5637	14.5474
0.0074	0.6412	300	0.0081	-286.9739	-284.4362	0.4718	-2.5377	-28530.3730	-28785.6816	14.5532	14.6383
0.0061	0.8549	400	0.0073	-253.9528	-254.9697	0.5040	1.0169	-25583.7266	-25483.5762	13.4697	13.4536

Framework versions

• Transformers 4.45.1
• Pytorch 2.4.1+cu121
• Datasets 3.0.0
• Tokenizers 0.20.0